Dirac 2.5 Operations
Lead Research Organisation:
University of Edinburgh
Department Name: Sch of Physics and Astronomy
Abstract
Physicists across the astronomy, nuclear and particle physics communities are focussed
on understanding how the Universe works at a very fundamental level. The distance scales
with which they work vary by 50 orders of magnitude from the smallest distances probed
by experiments at the Large Hadron Collider, deep within the atomic
nucleus, to the largest scale galaxy clusters discovered out in space. The Science challenges,
however, are linked through questions such as: How did the Universe begin and how is it evolving?
and What are the fundamental constituents and fabric of the Universe and how do they interact?
Progress requires new astronomical observations and experimental data but also
new theoretical insights. Theoretical understanding comes increasingly from large-scale
computations that allow us to confront the consequences of our theories very accurately
with the data or allow us to interrogate the data in detail to extract information that has
impact on our theories. These computations test the fastest computers that we have and
push the boundaries of technology in this sector. They also provide an excellent
environment for training students in state-of-the-art techniques for code optimisation and
data mining and visualisation.
The DiRAC-2.5 project builds on the success of the DiRAC HPC facility and will provide the resources needed
to support cutting edge research during 2017 in all areas of science supported by STFC.
DiRAC-2.5 will provide maintain the existing DiRAC-2 services from April 2017, and also provide and increase in computational
resources at Durham, Cambridge and Leicester.
This grant will support the operation of the Edinburgh DiRAC services, which presently comprise
98384 operational computing cores serving around 80% of DiRAC computing cycles. The system is made up
from both the original 1.26PFlop/s DiRAC BlueGene/Q system and, following a recent transfer
to Edinburgh by STFC, six racks of the Hartree BlueJoule supercomputer.
The DiRAC project also will offer a team of three research software engineers who will help DiRAC researchers to ensure their scientific codes to extract
the best possible performance from the hardware components of the DiRAC clusters. These highly skilled programmers will
increase the effective computational power of the DiRAC facility during 2017.
on understanding how the Universe works at a very fundamental level. The distance scales
with which they work vary by 50 orders of magnitude from the smallest distances probed
by experiments at the Large Hadron Collider, deep within the atomic
nucleus, to the largest scale galaxy clusters discovered out in space. The Science challenges,
however, are linked through questions such as: How did the Universe begin and how is it evolving?
and What are the fundamental constituents and fabric of the Universe and how do they interact?
Progress requires new astronomical observations and experimental data but also
new theoretical insights. Theoretical understanding comes increasingly from large-scale
computations that allow us to confront the consequences of our theories very accurately
with the data or allow us to interrogate the data in detail to extract information that has
impact on our theories. These computations test the fastest computers that we have and
push the boundaries of technology in this sector. They also provide an excellent
environment for training students in state-of-the-art techniques for code optimisation and
data mining and visualisation.
The DiRAC-2.5 project builds on the success of the DiRAC HPC facility and will provide the resources needed
to support cutting edge research during 2017 in all areas of science supported by STFC.
DiRAC-2.5 will provide maintain the existing DiRAC-2 services from April 2017, and also provide and increase in computational
resources at Durham, Cambridge and Leicester.
This grant will support the operation of the Edinburgh DiRAC services, which presently comprise
98384 operational computing cores serving around 80% of DiRAC computing cycles. The system is made up
from both the original 1.26PFlop/s DiRAC BlueGene/Q system and, following a recent transfer
to Edinburgh by STFC, six racks of the Hartree BlueJoule supercomputer.
The DiRAC project also will offer a team of three research software engineers who will help DiRAC researchers to ensure their scientific codes to extract
the best possible performance from the hardware components of the DiRAC clusters. These highly skilled programmers will
increase the effective computational power of the DiRAC facility during 2017.
Planned Impact
The expected impact of the DiRAC 2.5 HPC facility is fully described in the attached pathways to impact document and includes:
1) Disseminating best practice in High Performance Computing software engineering throughout the theoretical Particle Physics, Astronomy and Nuclear physics communities in the UK as well as to industry partners.
2) Working on co-design projects with industry partners to improve future generations of hardware and software.
3) Development of new techniques in the area of High Performance Data Analytics which will benefit industry partners and researchers in other fields such as biomedicine, biology, engineering, economics and social science, and the natural environment who can use this new technology to improve research outcomes in their areas.
4) Share best practice on the design and operation of distributed HPC facilities with UK National e-Infrastructure partners.
5) Training of the next generation of research scientists of physical scientists to tackle problems effectively on state-of-the-art of High Performance Computing facilities. Such skills are much in demand from high-tech industry.
6) Engagement with the general public to promote interest in science, and to explain how our ability to solve complex problems using the latest computer technology leads to new scientific capabilities/insights. Engagement of this kind also naturally encourages the uptake of STEM subjects in schools.
1) Disseminating best practice in High Performance Computing software engineering throughout the theoretical Particle Physics, Astronomy and Nuclear physics communities in the UK as well as to industry partners.
2) Working on co-design projects with industry partners to improve future generations of hardware and software.
3) Development of new techniques in the area of High Performance Data Analytics which will benefit industry partners and researchers in other fields such as biomedicine, biology, engineering, economics and social science, and the natural environment who can use this new technology to improve research outcomes in their areas.
4) Share best practice on the design and operation of distributed HPC facilities with UK National e-Infrastructure partners.
5) Training of the next generation of research scientists of physical scientists to tackle problems effectively on state-of-the-art of High Performance Computing facilities. Such skills are much in demand from high-tech industry.
6) Engagement with the general public to promote interest in science, and to explain how our ability to solve complex problems using the latest computer technology leads to new scientific capabilities/insights. Engagement of this kind also naturally encourages the uptake of STEM subjects in schools.
Publications
Katsianis A
(2021)
The specific star formation rate function at different mass scales and quenching: a comparison between cosmological models and SDSS
in Monthly Notices of the Royal Astronomical Society
Orkney M
(2021)
EDGE: two routes to dark matter core formation in ultra-faint dwarfs
in Monthly Notices of the Royal Astronomical Society
McAlpine S
(2022)
SIBELIUS-DARK: a galaxy catalogue of the local volume from a constrained realization simulation
in Monthly Notices of the Royal Astronomical Society
Sharma M
(2019)
The I?ea model of feedback-regulated galaxy formation
in Monthly Notices of the Royal Astronomical Society
Sirks E
(2022)
The effects of self-interacting dark matter on the stripping of galaxies that fall into clusters
in Monthly Notices of the Royal Astronomical Society
Fumagalli M
(2020)
Detecting neutral hydrogen at z ? 3 in large spectroscopic surveys of quasars
in Monthly Notices of the Royal Astronomical Society
Keating L
(2020)
Constraining the second half of reionization with the Ly ß forest
in Monthly Notices of the Royal Astronomical Society
Kirchschlager F
(2023)
Dust survival rates in clumps passing through the Cas A reverse shock - II. The impact of magnetic fields
in Monthly Notices of the Royal Astronomical Society
Hough R
(2023)
SIMBA - C : an updated chemical enrichment model for galactic chemical evolution in the SIMBA simulation
in Monthly Notices of the Royal Astronomical Society
Mitchell M
(2021)
A general framework to test gravity using galaxy clusters IV: cluster and halo properties in DGP gravity
in Monthly Notices of the Royal Astronomical Society
Irodotou D
(2022)
The effects of AGN feedback on the structural and dynamical properties of Milky Way-mass galaxies in cosmological simulations
in Monthly Notices of the Royal Astronomical Society
Hough R
(2023)
SIMBA - C : an updated chemical enrichment model for galactic chemical evolution in the SIMBA simulation
in Monthly Notices of the Royal Astronomical Society
Schaye J
(2023)
The FLAMINGO project: cosmological hydrodynamical simulations for large-scale structure and galaxy cluster surveys
in Monthly Notices of the Royal Astronomical Society
Wurster J
(2019)
Disc formation and fragmentation using radiative non-ideal magnetohydrodynamics
in Monthly Notices of the Royal Astronomical Society
Deason A
(2019)
The local high-velocity tail and the Galactic escape speed
in Monthly Notices of the Royal Astronomical Society
White S
(2020)
The globular cluster system of the Auriga simulations
in Monthly Notices of the Royal Astronomical Society
Santos-Santos I
(2020)
Baryonic clues to the puzzling diversity of dwarf galaxy rotation curves
in Monthly Notices of the Royal Astronomical Society
Brooks R
(2023)
The north-south asymmetry of the ALFALFA H i velocity width function
in Monthly Notices of the Royal Astronomical Society
Conaboy L
(2023)
Relative baryon-dark matter velocities in cosmological zoom simulations
in Monthly Notices of the Royal Astronomical Society
Deason A
(2023)
Unravelling the mass spectrum of destroyed dwarf galaxies with the metallicity distribution function
in Monthly Notices of the Royal Astronomical Society
Clarke C
(2020)
Forbidden line diagnostics of photoevaporative disc winds
in Monthly Notices of the Royal Astronomical Society
Hughes M
(2020)
The [a/Fe]-[Fe/H] relation in the E-MOSAICS simulations: its connection to the birth place of globular clusters and the fraction of globular cluster field stars in the bulge
in Monthly Notices of the Royal Astronomical Society
Harnois-Déraps J
(2023)
mglens : Modified gravity weak lensing simulations for emulation-based cosmological inference
in Monthly Notices of the Royal Astronomical Society
Ballabio G
(2023)
[O i ] 6300 Å emission as a probe of external photoevaporation of protoplanetary discs
in Monthly Notices of the Royal Astronomical Society
Arnold C
(2022)
forge : the f ( R )-gravity cosmic emulator project - I. Introduction and matter power spectrum emulator
in Monthly Notices of the Royal Astronomical Society
Glowacki M
(2022)
ASymba: H i global profile asymmetries in the simba simulation
in Monthly Notices of the Royal Astronomical Society
Gu Q
(2022)
The spatial distribution of satellites in galaxy clusters
in Monthly Notices of the Royal Astronomical Society
Kay S
(2020)
The intracluster light as a tracer of the total matter density distribution: a view from simulations
in Monthly Notices of the Royal Astronomical Society
Pfeffer J
(2022)
Using the EAGLE simulations to elucidate the origin of disc surface brightness profile breaks as a function of mass and environment
in Monthly Notices of the Royal Astronomical Society
Richings J
(2020)
Subhalo destruction in the Apostle and Auriga simulations
in Monthly Notices of the Royal Astronomical Society
Appleby S
(2020)
The impact of quenching on galaxy profiles in the simba simulation
in Monthly Notices of the Royal Astronomical Society
Qiao L
(2022)
The evolution of protoplanetary discs in star formation and feedback simulations
in Monthly Notices of the Royal Astronomical Society
Zheng Y
(2022)
Rapidly quenched galaxies in the Simba cosmological simulation and observations
in Monthly Notices of the Royal Astronomical Society
Lovell C
(2022)
An orientation bias in observations of submillimetre galaxies
in Monthly Notices of the Royal Astronomical Society
Chan K
(2022)
Single fluid versus multifluid: comparison between single-fluid and multifluid dust models for disc-planet interactions
in Monthly Notices of the Royal Astronomical Society
Dobbs C
(2022)
The formation of massive stellar clusters in converging galactic flows with photoionization
in Monthly Notices of the Royal Astronomical Society
Pakmor R
(2020)
The orbital phase space of contracted dark matter haloes
in Monthly Notices of the Royal Astronomical Society
Ruan C
(2022)
Towards an accurate model of small-scale redshift-space distortions in modified gravity
in Monthly Notices of the Royal Astronomical Society
Pettini M
(2020)
A bound on the 12C/13C ratio in near-pristine gas with ESPRESSO
in Monthly Notices of the Royal Astronomical Society
Mitchell P
(2020)
Galactic inflow and wind recycling rates in the eagle simulations
in Monthly Notices of the Royal Astronomical Society
Fancher J
(2023)
On the relative importance of shocks and self-gravity in modifying tidal disruption event debris streams
in Monthly Notices of the Royal Astronomical Society
Katz H
(2023)
The challenges of identifying Population III stars in the early Universe
in Monthly Notices of the Royal Astronomical Society
Beraldo e Silva L
(2020)
Geometric properties of galactic discs with clumpy episodes
in Monthly Notices of the Royal Astronomical Society
Haworth T
(2022)
An APEX search for carbon emission from NGC 1977 proplyds
in Monthly Notices of the Royal Astronomical Society
Evans T
(2022)
Observing EAGLE galaxies with JWST : predictions for Milky Way progenitors and their building blocks
in Monthly Notices of the Royal Astronomical Society
Ballabio G
(2023)
[O i ] 6300 Å emission as a probe of external photoevaporation of protoplanetary discs
in Monthly Notices of the Royal Astronomical Society
Satyavolu S
(2023)
The need for obscured supermassive black hole growth to explain quasar proximity zones in the epoch of reionization
in Monthly Notices of the Royal Astronomical Society
Orkney M
(2023)
EDGE: the shape of dark matter haloes in the faintest galaxies
in Monthly Notices of the Royal Astronomical Society
Davies J
(2020)
The quenching and morphological evolution of central galaxies is facilitated by the feedback-driven expulsion of circumgalactic gas
in Monthly Notices of the Royal Astronomical Society
Khachaturyants T
(2021)
How stars formed in warps settle into (and contaminate) thick discs
in Monthly Notices of the Royal Astronomical Society
Description | DiRAC 2.5 is a facility to support leading-edge computational astronomy and particle physics in the UK. This has resulted in over 500 peer-reviewed publications. |
Exploitation Route | Build on the scientific knowledge and computational techniques developed. |
Sectors | Digital/Communication/Information Technologies (including Software),Education |
Description | Intel IPAG QCD codesign project |
Organisation | Intel Corporation |
Department | Intel Corporation (Jones Farm) |
Country | United States |
Sector | Private |
PI Contribution | We have collaborated with Intel corporation since 2014 with $720k of total direct funding, starting initially as an Intel parallel computing centre, and expanding to direct close collaboration with Intel Pathfinding and Architecture Group. |
Collaborator Contribution | We have performed detailed optimisation of QCD codes (Wilson, Domain Wall, Staggered) on Intel many core architectures. We have investigated the memory system and interconnect performance, particularly on Intel's latest interconnect hardware called Omnipath. We found serious performance issues and worked with Intel to plan a solution and this has been verified and is available as beta software. It will reach general availability in the Intel MPI 2019 release, and allow threaded concurrent communications in MPI for the first time. A joint paper on the resolution to this was written with the Intel MPI team, and the application of the same QCD programming techniques to machine learning gradient reduction was applied in the paper to the Baidu Research all reduce library, demonstrating a 10x gain for this critical step in machine learning in clustered environments. We are also working with Intel verifying future architectures that will deliver the exascale performance in 2021. |
Impact | We have performed detailed optimisation of QCD codes (Wilson, Domain Wall, Staggered) on Intel many core architectures. We have investigated the memory system and interconnect performance, particularly on Intel's latest interconnect hardware called Omnipath. We found serious performance issues and worked with Intel to plan a solution and this has been verified and is available as beta software. It will reach general availability in the Intel MPI 2019 release, and allow threaded concurrent communications in MPI for the first time. A joint paper on the resolution to this was written with the Intel MPI team, and the application of the same QCD programming techniques to machine learning gradient reduction was applied in the paper to the Baidu Research all reduce library, demonstrating a 10x gain for this critical step in machine learning in clustered environments. This collaboration has been renewed annually in 2018, 2019, 2020. Two DiRAC RSE's were hired by Intel to work on the Turing collaboration. |
Start Year | 2016 |
Title | FP16-S7E8 MIXED PRECISION FOR DEEP LEARNING AND OTHER ALGORITHMS |
Description | We demonstrated that a new non-IEEE 16 bit floating point format is the optimal choice for machine learning training and proposed instructions. |
IP Reference | US20190042544 |
Protection | Patent application published |
Year Protection Granted | 2019 |
Licensed | Yes |
Impact | We demonstrated that a new non-IEEE 16 bit floating point format is the optimal choice for machine learning training and proposed instructions. Intel filed this with US patent office. This IP is owned by Intel under the terms of the Intel Turing strategic partnership contract. As a co-inventor I have been named on the patent application. The proposed format has been announced as planned for use in future Intel architectures. This collaboration with Turing emerged out of an investment in Edinburgh by Intel Pathfinding and Architecture Group in codesign with lattice gauge theory simulations. Intel hired DiRAC RSE's Kashyap and Lepper and placed them in Edinburgh to work with me on Machine Learning codesign through the Turing programme. |